From a16a3669273b4331fe5a0e4c7058c6e8df9d30b7 Mon Sep 17 00:00:00 2001 From: "Matthew Wilcox (Oracle)" Date: Tue, 14 May 2019 16:05:45 -0400 Subject: [PATCH] idr: Fix idr_get_next race with idr_remove commit 5c089fd0c73411f2170ab795c9ffc16718c7d007 upstream. If the entry is deleted from the IDR between the call to radix_tree_iter_find() and rcu_dereference_raw(), idr_get_next() will return NULL, which will end the iteration prematurely. We should instead continue to the next entry in the IDR. This only happens if the iteration is protected by the RCU lock. Most IDR users use a spinlock or semaphore to exclude simultaneous modifications. It was noticed once the PID allocator was converted to use the IDR, as it uses the RCU lock, but there may be other users elsewhere in the kernel. We can't use the normal pattern of calling radix_tree_deref_retry() (which catches both a retry entry in a leaf node and a node entry in the root) as the IDR supports storing entries which are unaligned, which will trigger an infinite loop if they are encountered. Instead, we have to explicitly check whether the entry is a retry entry. Fixes: 0a835c4f090a ("Reimplement IDR and IDA using the radix tree") Reported-by: Brendan Gregg Tested-by: Brendan Gregg Signed-off-by: Matthew Wilcox (Oracle) Signed-off-by: Greg Kroah-Hartman --- lib/idr.c | 15 +++++++-- tools/testing/radix-tree/idr-test.c | 52 +++++++++++++++++++++++++++++ 2 files changed, 65 insertions(+), 2 deletions(-) diff --git a/lib/idr.c b/lib/idr.c index fab2fd5bc326..61383564a6c5 100644 --- a/lib/idr.c +++ b/lib/idr.c @@ -231,11 +231,22 @@ void *idr_get_next(struct idr *idr, int *nextid) { struct radix_tree_iter iter; void __rcu **slot; + void *entry = NULL; unsigned long base = idr->idr_base; unsigned long id = *nextid; id = (id < base) ? 0 : id - base; - slot = radix_tree_iter_find(&idr->idr_rt, &iter, id); + radix_tree_for_each_slot(slot, &idr->idr_rt, &iter, id) { + entry = rcu_dereference_raw(*slot); + if (!entry) + continue; + if (!radix_tree_deref_retry(entry)) + break; + if (slot != (void *)&idr->idr_rt.rnode && + entry != (void *)RADIX_TREE_INTERNAL_NODE) + break; + slot = radix_tree_iter_retry(&iter); + } if (!slot) return NULL; id = iter.index + base; @@ -244,7 +255,7 @@ void *idr_get_next(struct idr *idr, int *nextid) return NULL; *nextid = id; - return rcu_dereference_raw(*slot); + return entry; } EXPORT_SYMBOL(idr_get_next); diff --git a/tools/testing/radix-tree/idr-test.c b/tools/testing/radix-tree/idr-test.c index 321ba92c70d2..235eef71f3d1 100644 --- a/tools/testing/radix-tree/idr-test.c +++ b/tools/testing/radix-tree/idr-test.c @@ -227,6 +227,57 @@ void idr_u32_test(int base) idr_u32_test1(&idr, 0xffffffff); } +static inline void *idr_mk_value(unsigned long v) +{ + BUG_ON((long)v < 0); + return (void *)((v & 1) | 2 | (v << 1)); +} + +DEFINE_IDR(find_idr); + +static void *idr_throbber(void *arg) +{ + time_t start = time(NULL); + int id = *(int *)arg; + + rcu_register_thread(); + do { + idr_alloc(&find_idr, idr_mk_value(id), id, id + 1, GFP_KERNEL); + idr_remove(&find_idr, id); + } while (time(NULL) < start + 10); + rcu_unregister_thread(); + + return NULL; +} + +void idr_find_test_1(int anchor_id, int throbber_id) +{ + pthread_t throbber; + time_t start = time(NULL); + + pthread_create(&throbber, NULL, idr_throbber, &throbber_id); + + BUG_ON(idr_alloc(&find_idr, idr_mk_value(anchor_id), anchor_id, + anchor_id + 1, GFP_KERNEL) != anchor_id); + + do { + int id = 0; + void *entry = idr_get_next(&find_idr, &id); + BUG_ON(entry != idr_mk_value(id)); + } while (time(NULL) < start + 11); + + pthread_join(throbber, NULL); + + idr_remove(&find_idr, anchor_id); + BUG_ON(!idr_is_empty(&find_idr)); +} + +void idr_find_test(void) +{ + idr_find_test_1(100000, 0); + idr_find_test_1(0, 100000); +} + void idr_checks(void) { unsigned long i; @@ -307,6 +358,7 @@ void idr_checks(void) idr_u32_test(4); idr_u32_test(1); idr_u32_test(0); + idr_find_test(); } #define module_init(x)